340 likes | 521 Views
TCP improvements for Data Center Networks -- Tanmoy Das and Krishna M. Sivalingam. B99705021 李奕德. outline. Abstract Intro ECN in DCTCP TDCTCP Performance evaluation conclusion. Abstract. TCP does not fit in datacenter DCTCP may lower throughput due to small buffer size
E N D
TCP improvements for Data Center Networks--Tanmoy Das and Krishna M. Sivalingam B99705021 李奕德
outline Abstract Intro ECN in DCTCP TDCTCP Performance evaluation conclusion
Abstract TCP does not fit in datacenter DCTCP may lower throughput due to small buffer size Improved version of DCTCP (called TDCTCP) Compare to existed algorithm (DCTCP, TCPNewReno) TDCTCP may have slightly higher delay but have much better throughput in general
outline Abstract Intro ECN in DCTCP TDCTCP Performance evaluation conclusion
introduction Data flow in datacenter: large data flowrequire high throughput small data flowrequire low latency Incastproblem
introduction TCP:provide : reliable, ordered byte streamdoes not provide: high throughput with “simultaneously low delay” Other similar solutions: DCTCP, TCPNewReno, TDCTCP Tested under OMNeT++ simulator
outline Abstract Intro ECN in DCTCP TDCTCP Performance evaluation conclusion
ECN in DCTCP ECN mechanism 2-bit used to describe current situation 00: Non ECN-Capable Transport — Non-ECT 10: ECN Capable Transport — ECT(0) 01: ECN Capable Transport — ECT(1) 11: Congestion Encountered — CE
outline Abstract Intro ECN in DCTCP TDCTCP Performance evaluation conclusion
TDCTCP A. Modification of Congestion Avoidance B. Resettingα after Delayed ACK timeout C. Dynamic Delayed ACK timeout calculation
TDCTCP-A part Modification of Congestion Avoidance α = fraction of marked packets in one congestion window Indicates current congestion level MSS = Maximum Segment Size Indicates the size of data that can be sent
TDCTCP-B part Resettingα after Delayed ACK timeout delayed ACK timeout: use in TCP to reduce ACKs send to the sender When ACK timeout occur: α is not updated Old α remain high and block increment of window size α is reset to 0 after every delayed ACK timeout
TDCTCP-B part Set α to 0 when this happen
TDCTCP-C part Dynamic Delayed ACK timeout calculation DCTCP: small buffer = small congestion window Congestion window reduce to 1, causing ACK timeouts
TDCTCP-C part Packet arrival follows an exponential distribution Packet loss probability in the network is small
TDCTCP Spend less time in ACK timeout Low variance in window size Use 10 flows to demonstrate
outline Abstract Intro ECN in DCTCP TDCTCP Performance evaluation conclusion
Performance evaluation Throughput Fairness Delay Queue length Variation in Delay Variation in Throughput
Performance evaluation Environment
Performance evaluation Throughput - single bottleneck, 1Gbps Better performance than DCTCP in general
Performance evaluation Throughput - single bottleneck, 10Gbps Better than DCTCP under smaller K Provide same throughput as TCPNewReno in early stages
Performance evaluation Throughput – multi-bottleneck, 10Gbps Better than DCTCP under smaller K Provide same throughput as TCPNewReno in early stages
Performance evaluation Fairness- single bottleneck, measure in JFI better fairness in every scenario
Performance evaluation Fairness- multi-bottleneck, measure in JFI
Performance evaluation Delay- single bottleneck , 10 Gbps TCPNewReno is good except high delay
Performance evaluation Delay- multi-bottleneck , 10 Gbps
Performance evaluation Queue length TDCTCP is slightly longer than DCTCP
Performance evaluation Variation in Delay
Performance evaluation Variation in throughput
outline Abstract Intro ECN in DCTCP TDCTCP Performance evaluation conclusion
conclusion Modified DCTCP => TDCTCP 15% higher throughput than DCTCP improved fairness compare to DCTCP provides more stable throughput queue length is slightly more than that of DCTCP at 10Gbps delay is slightly higher than that of DCTCP
The endthank you for your attention Any questions??